Skip to content. | Skip to navigation

Personal tools

Navigation

You are here: Home / Universities / McGill University / Refining access to health information using a novel indexing approach.

Refining access to health information using a novel indexing approach.

here is the summary of the research we are proposing
Title: Refining access to health information using a novel indexing approach.

Dawes M,  McKibbon  A. Bartlett J , Nie JY, Grad,R,  Pluye,P, Shea, L

Summary of Research Proposal – submitted for funding
The present research proposal aims to improve the performance of information retrieval technology for health professionals, and ultimately enhance traditional indexing systems by providing a new conceptual approach applicable across not only health but also all information sources. The information requirements of industry have expanded rapidly as the complexity of manufacturing and service structures have increased. For health care professionals the increase in information has been particularly evident. The information retrieval systems of many industries and professions have managed to deal with a large amount of information. However, the concepts employed are not different from general search engines. While online generic search engines such as Google aim to meet the needs of lay users, indexing systems satisfy the needs of traditional users, i.e., researchers and librarians. Both these systems rarely satisfy the needs of clinical (physicians, nurses, and other health professional) users. As a result people retrieving information for a specific question are now faced with search results that display too much information or irrelevant information.
The National Library of Medicine (NLM) has 16 million MEDLINE journal article references and abstracts. Each MEDLINE citation is indexed, both manually and automatically. The most important of these is the controlled vocabulary terms assigned by human indexers. NLM’s controlled vocabulary thesaurus, Medical Subject Headings (MeSH), contains approximately 23,000 descriptors arranged in a hierarchical structure. However the difficulty of information retrieval remains so severe that several countries have set up electronic web based question answering systems run by librarians to help clinicians find information.
A structured approach to question formulation for professionals has been developed that identifies the key elements of Patient/Population/Problem, Exposure/Intervention, Comparison, Outcome, Duration, Results (PECODR). We believe that using the PECODR framework to enhance the current indexing of the NLM would greatly improve health professional searches.
Our proposed research will first develop a controlled vocabulary for the PECODR elements from existing and developed taxonomies. We will then employ this vocabulary, representing these elements, to index a clinical important subset of the medical database (MEDLINE).  Finally we will test this system in answering questions compared to the current search and indexing system.
The first phase of this research will be to identify existing taxonomy for the PECODR elements. In the likely situation of some PECODR elements not having a validated taxonomy we will generate these and validate them using a Delphi approach. We will enhance the clinical retrieval technology by fine-tuning the index using the taxonomy to develop a new controlled vocabulary. With their rules, Demner Fushman’s group has been able to recognize around 80% of the PECODR elements in MEDLINE abstracts. The traditional named entity recognition can be coupled with domain taxonomy to recognize domain specific elements within the abstracts. We expect an increased success rate with this more sophisticated technique. The final phase will determine if the PECODR enhanced index is more effective than default search engines with respect to three outcomes:  Relevance, importance and impact. In the final stages of the project we will compare the use of the PECODR index comparing it with single term searches and combined term searches on the NLM search engine (PubMed) using the clinical queries search engine. Questions will be obtained from residents in our teaching units as well as faculty. Such an indexing system using the PECODR elements can be eventually linked directly to
an electronic medical record where the PECODR elements are already recorded and so could provide clinical support to the health professional at point of care in a completely novel approach.
Although this model for the improvement of information retrieval will test medical questions the principles discovered in this research can be applied to other industries and professions.

Document Actions

« October 2017 »
October
MoTuWeThFrSaSu
1
2345678
9101112131415
16171819202122
23242526272829
3031